Bahar QarabaqiAzar 19th, 1386

FC InferencingInitial information about the problem being

asserted into working memory.DatabaseSensorsUser

FC Inferencing (cont.)1. Scan the rules looking for ones whose

premises match the contents of the working memory.

2. Fire the rule which was found.3. Place its conclusion in the working memory.4. Until no additional rule fire, go to 1.

2. Fire the rule which was found… May locate several rules must decide Recognize-Resolve-Act cycle:

1. Scan the rules looking for ones whose premises match the contents of the working memory.

1’. Choose one rule to fire.2. Fire the rule which was found.3. Place its conclusion in the working memory.4. Until no additional rule fires, go to 1.

1’. Choose one rule to fire…Conflict resolutionSimplest:

Rules are examined in orderThe first rule is chosen

A common strategy:Each rule has a number which indicates its

priorityThe rule with the highest priority is chosen

…

Example 1:Pumping Station Diagnostic System

Block:a pump & a motorIncrease the water

pressure by 50 psiSensors:

line pressuremotor currents

Nominal values are available

An Event-Driven ES

Event-Driven vs. Conventional ESConventional ES

interacts with a user.Event-Driven ES

only becomes active when some special event occurs

Example 1–Problem Solving ApproachProblem solving strategy:

Fault detectionFault isolationFault diagnosis

Most ESs follow this sequence.Some also include fault response.Knowledge base is divided into various

sections

Example 1 – Fault Detection1. Numeric readings qualitative

descriptions2. Any fault low pressure 3. Faults propagate

Result: only need to monitor the final line pressure

Example 1 – Fault Detection (cont.)

Example 1 – Fault Detection (cont.)

Example 1 – Fault Isolation Why isolation?

By first identifying the faulty block, the system can concentrate its diagnostic effort on this single block.

Comparison of the block’s input pressure with its output pressure

Example 1 – Fault Diagnosis Which component is at fault?

Motor: low current Pump: no change in pressure Line: the block’s input pressure is less than

its output pressure

Example 1 – Fault Diagnosis (cont.)

Example 1 – Review Partitioned Rules:

Improve readability enhance maintenance

Intermediate Findings: Reports: what was observed and what the

system would look into next Intelligent Safety Net:

If the system is unable to determine the faulted component, knowing the general source may be valuable.

Example 1 – Review (cont.) Numeric Relationships:

49.99 is a low pressure! Solution: fuzzy logic

Specific Rules: Similar but separate rules for similar objects Solution: Example 2!

specific objects as variables

Example 2:Generalized Pumping Station Diagnostic System

Information about the structural relationship between components

Example 2 – Problem Solving ApproachProblem solving strategy:

General Fault DetectionGeneral Fault IsolationGeneral Fault DiagnosisGeneral Fault Response

Example 2 – General Fault Detection Fault Detection Heuristic:

If any line pressure drops below its nominal pressure, then you have a fault condition.

Used in RULE 1S

Example 2 – General Fault Detection

Example 2 – General Fault Isolation Fault Isolation Heuristic:

If you notice that a block’s input pressure is normal, but its output pressure is low, then the block may be faulty.

Example 2 – General Fault Diagnosis Which component is at fault?

Motor Problem Heuristic: A motor with a low current is suspect.

Pump Problem Heuristic: Pump problems usually result in no pressure

changes across the pump. Line Problem Heuristic:

When you see no problems with a block’s motor, but there is some increase in pressure across the block, there may be a leak in the output line.

Example 2 – General Fault Diagnosis

Example 2 – General Fault Response Purpose: replace the faulty component Only when the user has granted permission

Example 2 – Review Streamlining Rules:

A small number of general rules containing variables instead of a large set of rules.

Ease of Expansion: If additional objects added, only need to

assert their initial configuration information Stopping the System:

1. A common technique, but not the best: on their own

2. Force the system to stop

Example 2 – Review (cont.) Requesting Information:

1. Startup Rule:IF Get initial informationThen ASK …AND ASK …

2. When certain events occur1. As a part of a rule (our example)2. Demon Rule: the highest priority

IF ?Line pressure-status lowTHEN ASK shut down

Example 3:Train-Loading Expert System

Problem:Pack the passengers of various weights into a

series of train cars.Pack the persons by decreasing weight.Not exceed maximum weight capacity of train

car.Maximize the number of persons per train car

thus minimizing the number of cars needed.

ESs for Design ApplicationsDesign: configuring objects under a set of

constraintsConstraint:

1. Requirements: meet the design goal FC required: BC does not work in design

applications.

2. Methods: order of the steps1. Control Rules required:

IF <step> AND <heuristic> THEN <action>

2. Rule Ordering: simple to design, difficult to maintain3. Rule Priorities: difficult to maintain

Example 3–Problem Solving Approach

Example 3–Problem Solving Approach

Example 3–Problem Solving Approach

Example 3– Review Designed to Spec:

Final design met the stated specifications! Rules in Design Systems:

IF <step> AND <heuristic> THEN <action>